The calibration of Pepsin Digestibility

Nhi's picture

Is it impossibile to build the Pepsin Digestibility calibration?
I am making the calibration model of Fish Meal, it is seem good for proximate parameters (Moisture, Ash, Fiber, Protein, Fat), but for Pepsin, the result is very bad (RSQ, 1-VR <0). The range value is from 80 to 97%, the variation is good, so I don't know why. I'm using DS 2500 and WINISI to develop the calibration model.

ianm's picture

There is a lot of protein in fish meal. Pepsin is an enzyme, and also a protein. It is present in only a small quantity, and its absorbers would be similar to those of the bulk of protein, so NIRS is probably unable to differentiate clearly between the two type of protein, at least not well enough to enable a good calibration. The reproducibility of your method for determination of pepsin may also be a factor, unless it is very good.


(sent on behalf of Phil Williams by Ian Michael)

dwhopkins's picture

Hi NHitt,
I looked up the specificity of Pepsin to find out what peptide bonds are attacked, and Pepsin preferentially hydrolyses bonds next to the aromatic amino acids such as phe, tyr and trp, but not only these residues.  There are reports that various amino acids can be measured by NIR, but I have not been impressed by the claims.  So I think that it is asking too much of NIR calibrations to find these particular proteins rich in the aromatic amino acids to determine the susceptibility to Pepsin digestion.  I'd recommend giving it up, be happy with the proximate analysis of your feeds.
Best wishes,

Nhi's picture

Hi all,
Thanks for all advices. I have just worked in NIR for a short time, limited experience in NIR.  I see the infomation in some website, they have already the calibrations for Salt, amino acid, Fatty acid profile and Pepsin. I tried with these contituents, but it was really difficult to get the good calibration.
Thanks again!

gabiruth's picture

This reminds me of an old mistake I did in my earliest days in NIR back in 1998. I tried to measure very very low levels of some protein in a medium that contained other proteins
I learned the hard lesson that proteins are usually quite long chains containing the same amino acids even if possibly in different proportions and different chain locations
Despite that the spectral differences between an enzyme protein and all other proteins that are present in much much larger concentrations will be so small that NIR wouldn't be able to pick them up
Now let us understand what is the difference between 80 to 97% - it is a difference between 80% of nominal "level" and 97% of the nominal, it isn't a difference between 80 to 97% of enzyme in the matrix. Therefore we are probaby talking about concentrations in the below PPM level to further below PPM level. In such case Dave Hopking is absolutley right and I would probbaly not spend the time to try
Sometimes it could be possibele that some compounds in larger levels are present in the product and their levels are very stronlgy related to the level of the enzyme. If this condiion prevails, there is a possibility to fairly well estimate the enzyme level indirectly through the spectral changes of these compounds. The main problem in such a situation is that if for some reason this condition ceases to exist, the calibration will be useles.
Overall I have very litle faith in claims such this website claims that they have calibration for prbably micrograms of an enzyme
The situation with salt is somewhat similar, but more reliable because salt doesn't have its own spectrum, but being in ionic state and maybe electrically "tied" to protein polar points it exerts a specific measurable effect on the spectrum. The salt however is in higher levels and it ismpact on the spectrum is quite evident and measurable
I hope this is of help to you
Gabi Levin, Ph.D

khaled's picture

Hi to all,
Just to share my experience that previously we have success made calibrations of Pepsin Digestibility in Fish Meal with 4000 spectrum and we get RSQ/1-VR 0.8 -0.85.
RMSECV was about 1.1, and the instrument was FOSS 5000 model.
Pepsin digest is depend on the type of the fish meat which are used to make the fishmeal, and process parameter (such as heat) that can effet the protein become less digestible.
Both above factor are recorded in NIR spectrum that is why NIR can be used to predict The digestibility of the fish meal.

Dear Nihlt,
I would say to you that you have the chance to develop that on NIR but you need more spectrum and good laboratory to produce reliable data.
Good luck.

Khaled Sungkar.

ianm's picture

Before you attempt using these "factory-type" calibrations be sure to verify the reproducibility of all of your reference methods, and of the quality of the spectra that you will obtain from the material for which you are planning to use your instrument.

(posted by Ian Michael on behalf of Phil Williams)

khaled's picture

Agree with Ian, reproducibility of the reference method play the critical part.
Pepsin digestibilty lab analysis is not easy, it need lab technician with more experienced and good skills.

Best regards,
Khaled Sungkar

gabiruth's picture

Dear all
Reference method quality is a fundamental requirement. Without it there is no hope to achieve anything. I will be pleased to share another related fundamental requirement on reference data
If the range over which the calibration is done is R then the maximum allowed uncertainty in the reference values shall obey the following
U < Rx0.05 i.e. the U shall be smaller than 5% of the range. For a range of 80 to 97 i.e. 17% the U shall be less than 17x0.05=0.85. For better qualit I usually demand that U shall be smaller than 0.4xR and it helps. Therefore when you check the quality of the ref data be sure to check your quality against this requirement
I hope it helps
Gabi Levin, Ph.D.

gabiruth's picture

My apology for a typo error
I demand that U shall be smaller than 0.04xR of course
Gabi Levin